Running experiments on single entities¶

Note

The run_experiment script provides a quick and convenient way to execute any experiment from an actuator on a single point (entity) without the need to create a discoveryspace.

This is especially useful for rapid testing and debugging of actuator logic or experiment definitions.

Purpose¶

Quick Testing: Instantly run an experiment on a single point/entity to verify the behavior of the actuator or experiment.
No discoveryspace required: You do not need to create a discoveryspace or operation to use this tool.
Supports Local and Remote Execution: You can run experiments either locally or through an ado REST API endpoint.

Usage¶

run_experiment <point_file.yaml> [--remote <ENDPOINT>] [--timeout <SECONDS>] [--no-validate]

<point_file.yaml>: Path to a YAML file describing the point/entity and the experiments to run. This file should conform to the ado point definition schema.
--remote <ENDPOINT>: (Optional) If provided, the experiment will be executed through an ado REST API endpoint. If omitted, execution is local.
--timeout <SECONDS>: (Optional) Timeout for remote execution (default: 300 seconds).
--no-validate: (Optional) Skip entity validation before execution. This is useful if the experiment is not installed locally but is available remotely.

Example¶

An example point.yaml is given below. It defines the execution of the nevergrad_opt_3d_test_func from the custom_experiments actuator on the point x0=1, x1=3, x2=-1 (see optimizations with ado for more about this custom experiment).

The entity section contains a set of property/value pairs defining the point to run the experiment on. The experiments section is a list of experiment references detailing the experiment to run.

entity:
  x0: 1
  x1: 2
  x2: -1
experiments:
- actuatorIdentifier: custom_experiments
  experimentIdentifier: nevergrad_opt_3d_test_func

To execute the experiment, run

run_experiment example_point.yaml

Limitations & Notes¶

Single Point Only: This tool is designed for running experiments on a single point/entity.
For evaluating multiple points or running large-scale experiments, use a discoveryspace and the standard ado workflow.
No Metastore Tracking: The results from run_experiment are not tracked or stored in the ado metastore.
This means results are ephemeral and only available in the console output.
For Development & Debugging: This script is best suited for actuator/experiment development, debugging, or quick checks — not for production or persistent experiment tracking.

When to Use¶

Testing new actuator or experiment implementations.
Debugging experiment logic or entity validation.
Quickly verifying actuator installation and configuration.